The Food-101 dataset, a comprehensive collection of 101,000 food images across 101 categories, presents a challenging benchmark for image classification due to its intentionally noisy and mislabeled training data. This study uses the ResNet-50 architecture, a deep convolutional neural network known for its residual learning capability, to classify food images into binary categories: sweet and savory. Extensive preprocessing, including data cleansing, augmentation, and dimensionality reduction using Principal Component Analysis (PCA), was employed to prepare the dataset. In this study, we compared ResNet-50 with other models, such as VGG-16 combined with logistic regression, to understand the performance of different models on our dataset. The ResNet-50 model achieves the highest classification accuracy of 93.36%, showcasing ResNet-50 potential in fine-grained food image classification tasks.

Navigate to Specific Sections of the Project

Introduction

Data Preparation

Exploratory Data Analysis (EDA)

Classifier Models

Experimental Results

References

https://github.com/mohitsarin-tamu/ECEN-758-Project-Food-101

Developed By:

• Dorian Satuluri
• Mohit Sarin
• Sharvani Ramineni
• Vinodheni Ramasrinivasan
• Pavan Santosh